Machine Learning-Guided Production of a Self-Nanoemulsifying System for Delivery of Anacardic Acid

Bioactive molecules from plants remain an important source of drug candidates [1]. However, these have poor in vivo performance due to low water solubility leading to inadequate distribution [2,3,4], which can be counteracted by oil-in-water nanoemulsion drug delivery systems [5]. A limitation preventing the widespread adoption of nanoemulsions is the expensive, time-consuming iterative development process. In this study, we developed a nanoemulsion design by machine learning (ML). We retrieved average particle size and polydispersity index (PDI) data associated to nanoemulsion composition to construct a dataset from literature. A predictive ML model was used to identify improved self-nanoemulsifying systems including olive oil as base and combinations of Tween 20, Tween 80, glycerol, and soy lecithin. The predictive power of the model was determined by DLS. The nanoemulsions were loaded with pure anacardic acid. Encapsulation efficiency (EE%) was measured by HPLC, and the cytotoxic activity was evaluated on HEPG2, a human hepatic cancer cell line, and HEK-293, a normal-like human embryonic kidney cell line. The model’s accuracy was 81%. The best-performing formulation was 10% olive oil, 60% Tween 20, and 30% glycerol, exhibiting average particle size of 162.8±26 nm, a PDI of 0.234±0.03, and full EE%. The naked nanoemulsion presented no toxicity in HEK-293 but exerted an inhibitory effect on HEPG2 (IC50 20±1.07 μM). Moreover, loading the solution into the nanoemulsion increased the cytotoxic effect on HEPG2 in comparison to the naked nanoemulsion, and free anacardic acid, yielding an IC50 value of 12.4±0.3 μM. These results suggest that the formulation identified by the model was a successful carrier of the compound. This study presents a proof of concept on how ML can reduce the development pipeline of nanoemulsified drug delivery systems.